ScaleIO is a potential game-changer, disrupting the way how companies independent of their size can build new elastic Hyper Converged Private Cloud Infrastructures and how businesses purchase storage systems in the future.
ScaleIO is a solution which enables service providers to build not only a Hyper Converged Cloud which scales from 3 to 1000s of nodes providing high performance at low latency, it provides as well Cloud Framework and hypervisor independency supporting VMware (ESXi), OpenStack (KVM), Linux and in the near future Microsoft.
ScaleIO’s innovative technology called ECS (elastic converged storage) is a software-only solution built for the elastic 3rd generation platforms. It’s Software-Defined-Storage, a virtual Scale-Out storage array built on x86 commodity HW using any combination of solid-state devices, PCIe flash cards and hard drives providing fully protected and shared block LUNs.
5 reasons why ScaleIO is a disruptive technology – based on a customer PoC in Switzerland:
- 1. Hardware independency and asymmetric cluster architecture
Business Value: Lowest possible CAPEX for the infrastructure HW at any time and highest flexibility to address all future business needs in the most efficient way.
ScaleIO runs on any x86 server HW with any storage media (HDD, SSD, PCIe Flash) technology. Because of its asymmetric cluster architecture you can mix different servers in size and performance which allows you to immediately use the latest available technology providing the best price/performance/capacity ratio at any time and not only on day one when you start to build your infrastructure. It provides a flexible extension of the cloud with compute and storage (Hyper Converged) or compute only (VM’s) or storage only nodes as requested by the business. In our test we used 5 smaller servers (1xSSD) and 2 bigger servers (3xSSD) in the same cluster and it worked just great. We performed all kind of functional tests like fail a node, fail a drive, fail the ScaleIO VM, create LUN, expand LUN, SnapShot and restore of a LUN, QoS limitation per LUN and client and everything worked as expected.
- 2. Scalability to 100s or 1000s of nodes with auto expand and rebalancing functionality
Business Value: Provides a single pool of resources, dynamically grows as the business grows, resulting in a higher HW utilization and lower operational costs.
Scalability is a direct result of the architecture and must be given from the beginning. It cannot be added to a solution later on like any other feature. So this is important to test. If it does not scale today it will most probably not scale tomorrow. ScaleIO stands for true Scale-Out. Or have you ever seen before a storage array with 995 nodes providing 923’000 IO/s running on Amazon Web Services EC2 server? Look at the screenshot below showing the outcome of a test which was performed by EMC SE Matt Cowger and the ScaleIO engineering team around Boaz Palgi (CEO) and Erez Webman (CTO).
Impressive, but it’s not only around the 1000 nodes. In our PoC we proved the capability to add online additional nodes without the need to do any reconfiguration or even migrations to make the capacity and the performance available to all new and existing servers. And what was the result? Everything works in a fully automated manner, the added capacity and performance is available to all clients and the cluster automatically becomes rebalanced in the background without any manual interactions.
- 3. High performance at low latency
Business Value: You can run more applications in the cloud with a better user experience on less HW resources.
High performance at low latency is key as it enables you to run almost all workloads in your cloud. We tested ScaleIO together with VMware and achieved:
- 7-node cluster test: 86’000 IO/s with only 11 SSDs, 7’800 IOPS per SSD in a random read test
- Single VM test: 25’000 IO/s 8k full random 50% r/w ratio and an average latency between 1 and 2 ms!
- A throughput of up to 900 MB/s per node (10 Gb network limit)
Performance is always something which is influenced by many different factors. These results are less about the absolute numbers (which in this case is not a ScaleIO limitation) it’s about the latency, the amount of IO/s in a realistic read/write mix achieved with the available resources. As a “neutral” Swiss EMC Systems Engineer I can just say: Wow! This is fast and a fantastic result!
- 4. Online HW lifecycle management – no migrations, no customer impact
Business Value: Enables new dimensions of scalability and a reduction of the operational costs by factors.
If you have to run hundreds or thousands of nodes with hundreds of Terabytes or even hundreds of Petabytes of storage you have to not only be able to add new nodes online you have to be able to remove at the same time the oldest nodes from the cluster without any appreciable impact to the applications. In other words ScaleIO solves the supply chain challenge of a big cloud infrastructure at scale which massively reduces the OPEX and of course the TCO of the infrastructure. We have successfully tested this capability during our PoC as well.
- 5. Multi-cloud-support and automation through APIs
Business Value: Customer choice, no Cloud Framework look-in and highest possible agility through automation
A true SDS solution should provide you the flexibility to work with different Cloud Frameworks and hypervisors. ScaleIO supports VMware (ESXi) which we successfully tested, OpenStack (KVM) and even Bare Metal Linux environments. And soon ScaleIO will support Microsoft as well. You are able to integrate ScaleIO into any Multi-Cloud-Management system through a powerful API. In these days we are in discussions to expand the ScaleIO tests to OpenStack including the automation via the Cinder API. I will keep you updated.
EMC ScaleIO SDS – this is not a vision, nor a technology outlook – we tested it in Switzerland and declared the results as “amazing” and “disruptive”! Does this sound interesting and disruptive to you as well?
Feedback is very welcome!